更新 function.py

function.py

@@ -1,57 +1,79 @@
-import numpy as np
+import numpy as np  
-
+  
-id = "202210423105"
+# 填写学号和姓名  
-name = "魏明帅"
+id = "202210423105"  
-
+name = "魏明帅"  
-def train_logreg(X, y, learning_rate=0.01, iterations=1000):
+  
-    m, n = X.shape
+def sigmoid(z):  
-    # 添加偏置项
+    return 1 / (1 + np.exp(-z))  
-    X = np.c_[np.ones((m, 1)), X]
+  
-    weights = np.zeros(n + 1)
+def train_logreg(X, y, learning_rate=0.01, iterations=1000):  
-    
+    """  
-    for _ in range(iterations):
+    对率回归训练函数。  
-        for i in range(m):
+  
-            x_i = X[i]
+    Args:  
-            y_i = y[i]
+    - X (numpy array): 特征数据。  
-            z = np.dot(x_i, weights)
+    - y (numpy array): 标签数据。  
-            h = 1 / (1 + np.exp(-z))
+    - learning_rate (float): 学习率。  
-            error = y_i - h
+    - iterations (int): 迭代次数。  
-            weights += learning_rate * error * x_i
+  
-    
+    Returns:  
-    return weights
+    - numpy array: 训练得到的权重。  
-
+    """  
-def predict_logreg(inX, weights):
+    m, n = X.shape  
-    z = np.dot(inX, weights)
+    weights = np.zeros(n)  
-    h = 1 / (1 + np.exp(-z))
+    bias = 0  
-    return 1 if h >= 0.5 else 0
+  
-
+    for i in range(iterations):  
-def load_data(filepath):
+        h = sigmoid(np.dot(X, weights) + bias)  
-    X = []
+        error = h - y  
-    y = []
+        weights -= learning_rate * (1 / m) * np.dot(X.T, error)  
-    with open(filepath, 'r') as f:
+        bias -= learning_rate * (1 / m) * np.sum(error)  
-        for line in f:
+  
-            values = line.strip().split()
+    return weights, bias  
-            # 跳过包含 '?' 的行
+  
-            if '?' in values:
+def predict_logreg(inX, weights, bias):  
-                continue
+    """  
-            # 将标签和特征转换为浮点数
+    对率回归预测函数。  
-            y.append(float(values[-1]))
+  
-            X.append([float(x) for x in values[:-1]])
+    Args:  
-    
+    - inX (numpy array): 输入待分类的样本。  
-    X_array = np.array(X)
+    - weights (numpy array): 训练得到的权重。  
-    y_array = np.array(y).reshape(-1, 1)
+    - bias (float): 训练得到的偏置项。  
-    
+  
-    # 特征缩放，处理标准差为零的特征
+    Returns:  
-    means = np.mean(X_array, axis=0)
+    - int: 预测的分类标签。  
-    stds = np.std(X_array, axis=0)
+    """  
-    X_scaled = (X_array - means) / np.where(stds == 0, 1, stds)
+    prob = sigmoid(np.dot(inX, weights) + bias)  
-    
+    return 1 if prob > 0.5 else 0  
-    return X_scaled, y_array
+  
-
+def load_data(filepath):  
-if __name__ == "__main__":
+    """  
-    X_train, y_train = load_data("./data/horse-colic.data")
+    从文件中加载数据，若不自行正确实现，则无法通过评测，不记录成绩。  
-    X_test, y_test = load_data("./data/horse-colic.test")
+  
-
+    Args:  
-    weights = train_logreg(X_train, y_train, learning_rate=0.01, iterations=1000)
+    - filepath (str): 数据文件路径。  
-    y_pred = predict_logreg(X_test[11], weights)
+  
+    Returns:  
+    - np.array: 特征数据。  
+    - np.array: 标签数据。  
+    """  
+    data = np.loadtxt(filepath, delimiter=',')  
+    X = data[:, :-1]  
+    y = data[:, -1]  
+    y = (y > 0).astype(int)  # 将标签转换为1和0  
+    return X, y  
+  
+if __name__ == "__main__":  
+    # 加载训练集  
+    X_train, y_train = load_data("./data/horse-colic.data")  
+  
+    # 加载测试集  
+    X_test, y_test = load_data("./data/horse-colic.test")  
+  
+    # 训练对率回归模型  
+    weights, bias = train_logreg(X_train, y_train, learning_rate=0.01, iterations=1000)  
+  
+    # 测试单个数据  
+    y_pred = predict_logreg(X_test[11], weights, bias)  
     print(f"预测结果: {y_pred}, 实际结果: {y_test[11]}")